Bug 25747

Please boot into the Fisher installer until you see the screen that asks you
what kind of installation/upgrade you want, switch to VC2, then enter

mknod /tmp/hda
fdisk -l /tmp/hda
mknod /tmp/hdb
fdisk -l /tmp/hdb

And send me the fdisk output.

*** Bug 25746 has been marked as a duplicate of this bug. ***

*** Bug 25626 has been marked as a duplicate of this bug. ***

We (Red Hat) should really fix this before next release

Additional info as requested, 

fdisk -l /tmp/hda:

kernel version 132096

Disk /tmp/hda: 255 heads, 63 sectors, 1216 cylinders
Units = cylinders of 16065 * 512 bytes

   Device Boot    Start       End    Blocks   Id  System
/tmp/hda1   *         1         3     16600+  83  Linux
Partition 1 does not end on cylinder boundary:
     phys=(32, 15, 63) should be (32, 254, 63)
/tmp/hda2             3      1217   9753912    5  Extended
Partition 2 does not end on cylinder boundary:
     phys=(1023, 15, 63) should be (1023, 254, 63)
/tmp/hda5             3      1184   9491296+  83  Linux


fdisk -l /tmp/hdb
kernel version 132096

Disk /tmp/hdb: 255 heads, 63 sectors, 1216 cylinders
Units = cylinders of 16065 * 512 bytes

   Device Boot    Start       End    Blocks   Id  System
/tmp/hdb1   *         1      1052   8446000+  83  Linux
Partition 1 does not end on cylinder boundary:
     phys=(1023, 15, 63) should be (1023, 254, 63)
/tmp/hdb2          1052      1217   1324512    5  Extended
Partition 2 does not end on cylinder boundary:
     phys=(1023, 15, 63) should be (1023, 254, 63)
/tmp/hdb5          1052      1085    262552+  83  Linux
/tmp/hdb6          1085      1117    262552+  83  Linux
/tmp/hdb7          1117      1150    262552+  82  Linux swap


Let me know if you want me to try anything else.

 Daniel

It does not die - it is just notifying you of a situation which needs addressing
before we can install on the system.

I have changed the message to add some more information to aid the user in
resolving the system configuration issues which are blocking the install.

Ok, die is perhaps the wrong word, the system detects an anomaly, 
provides an information box then gracefully reboots. This loops 
ad-infinitum as even the Server/Custom classes won't "correct" 
the partition layout.

Note the help notes:
 for Workstation says "removes any Linux-related partitions..."
 for Server says "removes ALL existing partitions on ALL drives..."
 for Custom says "it is up to you how disk space should be partitioned.."

Clearly it doesn't get as far as doing any of these!

I'll now modify the partitions myself.

 Daniel

Ok I think I spoke too soon - you are suggesting that for install classes that
were going to blow EVERYTHING away anyhow, we should offer option to zero the
partition table?

But for those classes the installation never gets far enough to blow anything
away.

I get the warning for hda and the option to Retry or Skip Drive. I've just
tried retry 30 times and I can't get passed it. So I Skip Drive and get
same for hdb. Then the final box say "An error has occurred - no valid
devices were found on which to create new filesystems. Please check 
your hardware for the cause of this problem." The only option is OK, 
which gracefully shuts down and restarts the system.

Perhaps therefore you have to add the option to zero the partition table,
or a note to switch to a VC and manually manipulate the partition table.
Either way the potential outcome is the loss of original data.
The warning for Server installation certainly seems to suggest to me that 
EVERYTHING on all installed disks is going to be deleted.

Will address in next release.

Next release means Wolverine or what? I tried installing on a Compaq Deskpro where I had previously installed beta3 (ws setup), together with NT4 and I 
got the same message. Unfortunately, I have only one disk, so installer aborts.

This REALLY IS buggy and you REALLY SHOULD fix it, rather than close it
saying WONTFIX.  Learn something from the Cygnus team -- we built a bug
tracking system that wouldn't let engineers or tech support staff close
out a bug until the CUSTOMER agreed that it was fixed, or didn't need to be.

I have a RedHat 5.2 system which I'm trying to upgrade to 7.1 using the
CD.  When the 5.2 system was installed, due to lots of crazy machinations
due to the 5.2 kernel not being able to handle C/H/S for 16GB disk, my
partitions were created on what the disk label thinks are "not cylinder
boundaries".

When I try to upgrade using RH 7.1, I get to the "Install or Upgrade"
screen and click Upgrade and Next.  Then it pops up a box saying that
some of my partitions don't end on a cylinder boundary, and gives me
NO OPTION except to click "Skip Drive".  Then it immediately pops up
another box saying "An error has occurred - no valid devices were found
on which to create new file systems."  Again, it gives me NO OPTION
except to click OK.  (I've tried going over to the Back button and it
DOES NOT WORK when facing one of these stupid mode-full dialog boxes.)
As soon as I click OK, the whole Anaconda terminates and the system
reboots.  It appears to be DESIGNED TO DO THIS.  That's what we are
all complaining about in this bug report.

So, there are several things to fix in here:

  *  Don't make the system reboot if you can't find any disk drives!
     Just let the sysadmin pick Back and try something else.

  *  Don't complain if partitions are supposedly "misaligned".  The
     GNU Parted (version 1.4.18) manual says it quite clearly:  there
     are no constraints on where you can put Linux partitions, they
     can go at any sector on the disk, because Linux uses LBA mode.
     Lilo will also use LBA mode if you give the "linear" option in
     /etc/lilo.conf.  The whole concept of "cylinders" is meaningless
     in every modern hard disk; to fit more data on the disk, they put more
     sectors on the outer tracks than on the inner tracks, and they
     sure don't tell US where those boundaries are.  Not only Anaconda,
     but the rest of the Linux disk utilities should ignore "cylinders"
     and "heads" and just use linear addresses.